[dekhn]

Yes but in those situations, you typically can at best find associations between variables and outcomes. We really want evidence of causality, although it sort of depends on how you interpret the precautionary principle.